My first ARM/LPC design, critic requested.

[Jeroen3]

Read here: http://learnembedded.blogspot.nl/2009/11/embedded-artists-lpc-2148-usb.html

USB powered applications require a FET if the application can use more than one unit load. A unit load is 100 mA.
If your application requires the full 500 mA available on standard ports, the chip must ask for it. And if it's allowed it can safely turn on the heavier loads.
Strictly implemented hosts might even turn off the power to a downstream port when the device is using more than the assigned unit loads.
This is actually very rare since the "sleep" state power usage of USB devices is also limited to a few mA. Overload protection isn't that rare, but often cheaped away.
But with bus-powered USB hubs this is a trap to be aware of, since all nested devices share a single downstream port.
That is also why smartphones don't charge on USB ports where the host does not enumerate. Unless there is a specific voltage on d- and d+ like the chargers do.

[westfw]

If your application requires the full 500 mA available on standard ports, the chip must ask for it.

That would be a lot more believable if I'd ever seen a USB host actually implement it.
Instead, there are any number of devices (chargers, fans, CUP WARMERS!) that happily draw 100mA without even containing the intelligence needed to "request" full power.

[paulie]

USB ports on my desktop and netbook are a main source of power for nearly all my 5v and 3.3v projects. Many like serial WIFI modules and TFT displays suck 300-400ma or more no problem just using the outside pins. With direct short I will get an XP popup warning. Like with many things there does seem to be quite a gap between theory and practice.

[nctnico]

That depends very much on the implementation of the USB port. Some ports are limited to 100mA unless the device has negiotiated more current. Other ports are just limited to a guaranteed 500mA current.

[SirNick]

Like with many things there does seem to be quite a gap between theory and practice.

Yep, because the theory is close to unreasonable.  You basically need a USB controller IC that can selectively disable the PSU to everything else in the product to comply with the sleep current draw requirements.

[zapta]

Just completed the layout. It's a 0.5mm pitch IC with full ground tab below so fanout was a struggle but the Eagle auto router helped.

https://github.com/zapta/arm/blob/master/eagle/ezlpc-proto-board.pdf?raw=true



Parts will arrive from Digikey in a few days. I will verify the footprints and will send the gerbers out.

[Jeroen3]

Your SWD cable will need to go over the buttons since the notch (uneven side) is most often found at the opposite side of the cable at an IDC plug.
The assembler house might reject your pcb if you put untented via's too close to your pads (see U1).
I'm also not seeing any large tantalums of at least 4u7. (they come in 1210) Neither a ferrite, which you will need for clean USB powered analogs.

All other placing and routing looks acceptable. Hope it works first try 

[nctnico]

I would have optimised the schematic more so the bottom could have a solid ground plane. I don't know the via size but I think they can be made smaller. I would have routed this board manually for maximum ground plane coverage.

[paulie]

Actually that board is almost exactly the one I would have done if that chip were my target. Pro-mini-ARM! Zapta, do you have an estimate for parts cost yet? Did you use OSH? Not only are they fastest turn around but lowest cost in USA. I think even cheaper if you choose open design on his site which allows others to order direct.

[zapta]

Your SWD cable will need to go over the buttons since the notch (uneven side) is most often found at the opposite side of the cable at an IDC plug.

I will check it again. I am using an NXP Link 1 debugger and I designed this orientation to be natural when using this board on a solderless breadboard.

The assembler house might reject your pcb if you put untented via's too close to your pads (see U1).

These vias are covered with solder mask. Does answer this concern? If not, what should be the clearance for same and for different signal vias?  (BTW, the assembler house is me).

I'm also not seeing any large tantalums of at least 4u7. (they come in 1210) Neither a ferrite, which you will need for clean USB powered analogs.

I am going after an example from Embedded Artists (semi official NXP  board).  They use non polarized 2.2u on both sides of the LDO and 0.1u + 0.01u on each of the two VDD pins (I left one capacitor per VDD pin).

As for the  frrite, the sample board had one on the USB +5V but I dropped it for simplicity. This is a hobby project, no need compliance, just working well in practical scenarios.

 Is it good enough?

All other placing and routing looks acceptable. Hope it works first try 

My boards typically do, will see about this one ;-)

[zapta]

Actually that board is almost exactly the one I would have done if that chip were my target. Pro-mini-ARM! Zapta, do you have an estimate for parts cost yet? Did you use OSH? Not only are they fastest turn around but lowest cost in USA. I think even cheaper if you choose open design on his site which allows others to order direct.

Hi Paulie, Haven't decide yet if to send the prototypes to elecrow (1mm, HASL, with metal stencil) or oshpark (1.6mm, ENIG, plastic stencil from oshstencils.com).  In any case, I am working against a public repository so all the files are available https://github.com/zapta/arm

I ordered parts for three boards. Will be able to send you one assembled or just the PCB, as you prefer. Hopefully you will find it useful and will be able to improve on it (hardware and software).  The main component is the 568-9586-ND,  $4.54 single, $2.27 at 1k.  I didn't try to optimize cost too much.  This is a non commercial project for me and people are free to do with it whatever they want.

BTW, one goal of this project is to make the entire process of developing a custom app simple. It include a single LPCXpresso project with all the necessary libraries so installation should be a breeze (and OS agnostic).    This is the current hello world + blinky https://github.com/zapta/arm/blob/master/lpcxpresso/hello_world/src/main.cpp  , it is outlined after the Arduino setup/loop model and include a non blocking delay using the passive_timer class.

[zapta]

Actually that board is almost exactly the one I would have done if that chip were my target. Pro-mini-ARM! Zapta, do you have an estimate for parts cost yet?

I may adopt that name, ARM Pro Mini  ;-)  Currently I call it ezLPC.

Just uploaded the BOM here

https://github.com/zapta/arm/blob/master/eagle/ezlpc-proto-bom.pdf?raw=true

I harmonized resistor values as much as possible. Some parts can have less expensive versions, for example the LDO.

[nctnico]

I'm also not seeing any large tantalums of at least 4u7.

Please no tantalums. Just use X5R or X7R MLCC for such capacitances.

[paulie]

Will be able to send you one assembled or just the PCB, as you prefer.

Actually I'm inclined toward something in between. Board with chip only would allow me to experiment with minimum parts count by progressively installing components. I.E. the regulator and associated components not required because virtually all my supplies are 3.3v these days. I'd also like to experiment to see what parts of the USB circuit are really needed along with many of the caps including crystal and bypass.

BTW there don't seem to be any Gerbers in your repository. Many who don't run Eagle including Kicad users like me are left out in the cold. Another reason OSHs open file deal might be a good choice. That and your board would only cost about a buck or two.

Solder template? Real Men use toothpick and hair dryer.

[Jeroen3]

Please no tantalums. Just use X5R or X7R MLCC for such capacitances.

Do you have any motivation with that? Since tantalums are a nice compromise between size of electrolytic and the price of ceramic.
And since you must consider USB the most dirty (noise!) source with a voltage between 4.5 and 5.5 a 2.2 uF cap might be a bit low-ish.
But hey, it's a devboard...

[paulie]

From what I can tell tantalums are a holdover from the days of aluminum cans, half watt resistors, and ceramic disk. Back when monolithics were through hole and SMD strange and fearful territory. I've noticed that for small circuits like modern RC speed controllers and Adruino modules the tantalums are disappearing in favor of cheaper, smaller, and lower ESR "chip" capacitors. Again hit counts on Ebay are a hint for component popularity and availability.

Not to mention a major cause of infant mortality is polarized caps installed backwards. Not a problem with monolithics.

[Kjelt]

AFAIK tantalums have their bad reputation from their old predecessors that had a tendency to explode and have limited life.
The present generation tantalums are pretty good AFAIK and I see them often in quality A brand products, so wonder why they would do that?

[paulie]

So your theory is the new tantalums work OK when installed backwards? This conflicts with my own recent personal experience.

I'm not sure about "A brand" products because gullible is not my middle name but the other 99% out there are switching over.

[zapta]

Will be able to send you one assembled or just the PCB, as you prefer.

Actually I'm inclined toward something in between. Board with chip only would allow me to experiment with minimum parts count by progressively installing components. I.E. the regulator and associated components not required because virtually all my supplies are 3.3v these days. I'd also like to experiment to see what parts of the USB circuit are really needed along with many of the caps including crystal and bypass.

BTW there don't seem to be any Gerbers in your repository. Many who don't run Eagle including Kicad users like me are left out in the cold. Another reason OSHs open file deal might be a good choice. That and your board would only cost about a buck or two.

Solder template? Real Men use toothpick and hair dryer.

I renamed the board to arm-mini-pro, thanks for the suggestion.  Also rotated the SWD connector (thanks  jeroen3).   Files are here

https://github.com/zapta/arm/tree/master/mini-pro/eagle

Will generate gerbers once I verify the footprints with actual components (ordered from Digikey).  BTW, OSHpark accepts eagle files directly and I often use it to view the file during layout.

Once it works I can send you a board with any combination of populated and populated components you like. It may be easier to remove components that adding them.  Also, since I ordered spares just in case, there is no real cost for me.

[Kjelt]

So your theory is the new tantalums work OK when installed backwards?

Duh, Ofcourse not that is not what I wrote. Neither do any other polarized capacitors like elco,s by the way  .

[paulie]

Ooops... Sorry.I thought you were replying to my comment about backwards:

tantalums have their bad reputation from their old predecessors that had a tendency to explode and have limited life.

Now I see you just meant they used to do that on their own which is true. The technology did get much better with time and there was much less tendency to do that. Now they only explode and have limited life when reversed. Actually they just tend to get hot, crack, and sometimes glow. But I would personally avoid them in favor of the ceramic chips just for cost and better noise suppression anyway.

Also worth mentioning that you can get more capacitance in smaller space with chip caps. I just purchased several reels of 22uf chips in 0603 for a client. That's TWENTY TWO MICROFARADS, not 2.2! What tiny little buggers and less than 2 cents per.

[Kjelt]

Also worth mentioning that you can get more capacitance in smaller space with chip caps. I just purchased several reels of 22uf chips in 0603 for a client. That's TWENTY TWO MICROFARADS, not 2.2! What tiny little buggers and less than 2 cents per.

True just remember that mcc have that nasty voltage derating, so that a 2u2/6V3 mcc can have only 30% of that capacitance left at 5V, so you have to use a 2u2/16V or even 25V for it.
But I agree that mcc are the way to go esp. low capacitance (<=10uF).
Where I still sometimes use a tantalum is if I have very little room and still need high value capacitor for instance I have here 470uF/6V3 which have a 6x4mm SMD enclosure, I find that a very good uF/volume deal.

[SirNick]

I'm finding it hard to justify tantalums.  I looked at prices for input-side filter caps on a quickie design I did last weekend.  ~$1.20 in single quantity for a 22uF 35v MLCC in a 1210 package.  Somewhere in the $4 range for a bigger tantalum-polymer at the same capacity, and equivalent working voltage.  Of course they both get much cheaper at lower voltages, but this is pre-regulator, where I want to be as lenient as possible.  (I ended up settling for 22uF 25v MLCC, which was half the cost.)

When it gets too pricey to fit a MLCC, I just go directly to electrolytic.  I'm a little put off by tantalum temper tantrums WRT reflow solder damage and applied voltage thereafter.  At least when a 'lytic fails, it pops and makes confetti, rather than bursting into flame.  IIRC, the polymer variety aren't supposed to be quite as violent though.

[nctnico]

AFAIK tantalums have their bad reputation from their old predecessors that had a tendency to explode and have limited life.
The present generation tantalums are pretty good AFAIK and I see them often in quality A brand products, so wonder why they would do that?

Tantalums will still explode when mounted in reverse. And yes that happens even in an automated process. Another problem with tantalums is that you need to control the current slew rate into them. If the slew rate is too high they might go off. A tantalum is basically a thermite lance/bomb because they consist of a fuel and an oxidizer. If they go off they can burn right through a PCB making it necessary to replace the entire PCB.

Besides that MLCC is cheaper nowadays and has a much lower ESR. So why bother with tantalums. The only reason to use a tantalum is when you need a relatively large capacitance for which there is no MLCC alternative. Another reason could be that you need a slightly higher ESR due to a regulator not being stable with MLCC capacitors. Still the strong influence of voltage on capacitance makes it tricky to select the proper MLCC. The capacitance is usually highly overrated.

Bottom line: I never use tantalums in my designs unless there is no alternative.

[zapta]

Tantalum are blood capacitors

http://news.bbc.co.uk/2/hi/africa/1468772.stm